Commits
Author  Commit  Message  Labels  Comments  Approvals  Date 

Fix cumulative for distributions
Properly handle infinities and NaNs




Workaround for slowness of incompleteBeta
incompleteBeta p p 0.5 is very sow for large p




Describe how shlould cumulative be evaluated at infinities




6af937a
M

Merge branch 'master' of github.com:bos/statistics




Add case for sample which contain same elements




Disallow grids with single point.
Such grids don't make sense anyway.
Fix #37




Fix bug in dct for vectors with length 1
It have to be treate specially. Also index permutaions do not wotk
for vectors that aren't power of two. E.g. dct [1,1,1] will produce vector
with length 2 insterad of error. So check were move to the outermost functions
It in fact improves error messages
Fix #37




Inline wrapper function and give them change to fuse




Add unit test for DCT/IDCT for vector with length 1




Fix haddock




Fix edge case for F distribution
Following should hold:
> cumulative d +Infinity = 1




Avoid overflow of exponent in calculation of geometric mean
Fixes #38 by Alex Lang




Fix tiny typo




Update tags again




Unbreak test suite




Added tag 0.10.2.0 for changeset ed4e716bb512




Rebuild tags file



Add benchmark for FFT




Change semantics if dct_ and idct_
They now transform only real part of vector. Imaginary part is
discarded. Previous behavior was to transform imaginary part
of vector with some linear transformation which is not DCT/IDCT
Should fix #16




5441e92
M

Merge pull request #35 from 00tau/master
Documentation for the F distribution said T distribution





erratum: TDistribution should read FDistribution




Update changelog




Fix module list and dependencies for the test suite
Fixes #31




Adjust KDE estimate to new dct definition




Apply same change for CDF test




Fix test for quantile.
Select p more efficiently




Add test for integrtal of KDE




Check for samples with zero and one element
KDE is not defined for empty sample and automatically determined
range is not sensible for 1 element range




Fix both DCT and IDCT
Implementation follow FFTW3 convention which multiply definition
from wikipedia by two




Add C program to generate test cases for DCT


